Large-view-field optical remote sensing satellite in-orbit geometric calibration method based on multi-mode fusion

Abstract

The invention discloses a large-view-field optical remote sensing satellite geometric calibration method based on multi-state data fusion, and belongs to the field of optical remote sensing satellite geometric calibration. According to the invention, a high-precision correlation control model based on overlapping region constraint and calibration parameter transmission is established, through mutual iterative optimization of internal and external orientation elements, internal orientation element calibration parameter optimization is carried out by using multi-modal imaging parameters, and optimization is carried out step by step to a full field of view, so that high-precision inter-slice splicing of images is realized, and high-precision inter-slice splicing is realized. And based on the obtained full-view-field optimized interior orientation elements, establishing an exterior orientation element overall correction model through overlapping region connection points, performing multi-modal fusion on multiple times of exterior orientation element calibration parameters, completing exterior orientation element consistency correction, and finally realizing high-precision geometric calibration under a large-view-field imaging condition. According to the large-view-field optical remote sensing satellite in-orbit geometric calibration method based on multi-modal fusion, the problem that large-view-field geometric calibration cannot be completed at a time through a geometric calibration field is solved.

Description

Technical Field

[0001] The invention belongs to the field of optical remote sensing satellite geometric calibration, and in particular relates to a large-field optical remote sensing satellite geometric calibration method based on multi-state data fusion. Background Art

[0002] On-orbit geometric calibration is a key step in achieving high-precision geometric positioning of optical remote sensing satellites. The vibration environment during launch and the long-term space environment after entering orbit, the changes in the camera structure composed of multiple sensors, and the slow drift of star sensitivity will all lead to a decrease in the accuracy of satellite geometric positioning. Geometric calibration must be performed regularly after entering orbit and during on-orbit operation to update the calibration parameters to ensure the positioning accuracy of image products.

[0003] Traditional on-orbit geometric calibration of optical remote sensing satellites usually uses satellite imaging to obtain calibration field image data based on the geometric calibration field, establishes a geometric calibration model based on the probe pointing angle, and uses the high-precision DOM and DEM data of the geometric calibration field to match the control point image, extract high-precision control point pairs, and perform iterative calibration of the camera's internal and external orientation elements to complete the geometric calibration. The basic process is as follows: Figure 1 shown.

[0004] Compared with the traditional optical remote sensing satellite geometric calibration method, the present invention can obtain cloud-free data within the full field of view of the camera sensor at one time under narrow field of view conditions, and then carry out geometric calibration work based on a high-precision geometric calibration field. However, for high-resolution ultra-large field of view imaging camera sensors with a width of up to several hundred kilometers, it is impossible to obtain cloud-free data covering all slices within the full field of view at one time. It is necessary to use multiple imaging data for joint geometric calibration processing. However, when imaging at different times, there are differences in the measurement errors of the satellite's orbit, attitude, travel time, etc., which can easily cause inconsistency in the calibration of internal and external orientation elements, thereby leading to problems of inter-slice splicing image misalignment and geometric positioning deviation. Therefore, the present invention proposes an on-orbit geometric calibration method for large-field optical remote sensing satellites based on multi-state data fusion. Summary of the invention

[0005] (I) Purpose of the invention

[0006] The purpose of the present invention is to provide an on-orbit geometric calibration method for a large-field-of-view optical remote sensing satellite based on multi-state data fusion, so as to solve the problem of inconsistency in calibration of internal and external orientation elements caused by using multiple imaging data for geometric calibration, resulting in misalignment of inter-slice stitching images and geometric positioning deviation, thereby ultimately solving the problem of on-orbit geometric calibration of high-resolution and ultra-large-field-of-view imaging images and obtaining high-precision geometric positioning images.

[0007] (2) Technical solution

[0008] In order to achieve the above objectives and solve the above technical problems, the technical solutions of the present invention are as follows:

[0009] The on-orbit geometric calibration method of a large-field-of-view optical remote sensing satellite based on multi-modal fusion includes the following steps:

[0010] Step 1: Geometric calibration of single satellite observation data. Using high-precision geometric calibration field data as a reference, a geometric calibration model for cloud-free areas is constructed. By solving the geometric calibration model, the internal orientation elements and camera external orientation elements corresponding to each pixel in the sensor of a single cloud-free area of ​​camera imaging are obtained.

[0011] Step 2: Geometrically calibrate the cloud-free satellite observation data containing the overlapping area twice. Based on the internal orientation elements of the first calibration in the overlapping area, the geometric calibration model of single satellite observation data is used to calculate the camera external orientation elements corresponding to the second imaging sensor. Then, the camera external orientation elements are used to calibrate the internal orientation elements of the sensor without overlap in the second imaging.

[0012] Step 3: using the overlapping area between other secondary imaging data and the first or second imaging data, and using the method of step 2, obtain the exterior orientation elements of other secondary imaging data and the interior orientation elements of all other sensors.

[0013] Step 4: Based on the internal orientation elements acquired within the full field of view, the overall correction model of the external orientation elements is established by connecting the points in the overlapping area, and the calibration parameters of multiple external orientation elements are multimodally fused to complete the consistency correction of the external orientation elements. Through the mutual iterative optimization of the internal and external orientation elements, high-precision geometric calibration with a large field of view is completed.

[0014] (3) Effective income

[0015] 1. The in-orbit geometric calibration method for a large-field-of-view optical remote sensing satellite based on multi-modal fusion proposed in the present invention solves the problem that the large-field-of-view geometric calibration cannot be completed in one go through the geometric calibration field.

[0016] 2. The present invention establishes a high-precision correlation control model based on overlapping area constraints and calibration parameter transfer. Through mutual iterative optimization of internal and external orientation elements, the calibration parameters of the internal orientation elements are optimized using multimodal imaging parameters, and gradually optimized to the full field of view, thereby achieving high-precision inter-slice stitching of images. Based on the acquired full-field optimized internal orientation elements, an overall correction model of the external orientation elements is established through the overlapping area connection points, and the calibration parameters of multiple external orientation elements are multimodally fused to complete the consistency correction of the external orientation elements, ultimately achieving high-precision geometric calibration under large-field-of-view imaging conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Traditional satellite on-orbit geometry calibration process;

[0018] Figure 2A schematic diagram of the on-orbit geometric calibration process of a large-field-of-view optical remote sensing satellite according to the present invention;

[0019] Figure 3 Schematic diagram of internal and external element transfer. DETAILED DESCRIPTION

[0020] The present invention is explained and described in detail below with reference to the accompanying drawings.

[0021] The idea of ​​solving the problem of the present invention is generally described as follows: the present invention obtains geometric calibration field data multiple times, takes high-precision geometric calibration field data as a reference, and obtains external orientation elements and internal orientation elements covering the reference area through geometric calibration model solution; then establishes a high-precision correlation control model based on overlapping area constraints and calibration parameter transmission, optimizes the calibration parameters of the internal orientation elements using multimodal imaging parameters, and gradually optimizes to the full field of view, thereby achieving high-precision image inter-slice stitching; based on the obtained full-field optimized internal orientation elements, an overall correction model of the external orientation elements is established through the overlapping area connection points, and the calibration parameters of multiple external orientation elements are multimodally fused to complete the consistency correction of the external orientation elements. Through the mutual iterative optimization of internal and external orientation elements, large-field high-precision geometric calibration is completed.

[0022] The present invention is mainly divided into the following 4 steps:

[0023] Step 1: Geometric calibration of single satellite observation data. Using high-precision geometric calibration field data as a reference, a geometric calibration model for cloud-free areas is constructed. By solving the geometric calibration model, the internal orientation elements and camera external orientation elements corresponding to each pixel in the sensor of a single cloud-free area of ​​camera imaging are obtained.

[0024] Step 2: Geometrically calibrate the cloud-free satellite observation data containing the overlapping area twice. Based on the internal orientation elements of the first calibration in the overlapping area, the geometric calibration model of single satellite observation data is used to calculate the camera external orientation elements corresponding to the second imaging sensor. Then, the camera external orientation elements are used to calibrate the internal orientation elements of the sensor without overlap in the second imaging.

[0025] Step 3: using the overlapping area between other secondary imaging data and the first or second imaging data, and using the method of step 2, obtain the exterior orientation elements of other secondary imaging data and the interior orientation elements of all other sensors.

[0026] Step 4: Based on the internal orientation elements acquired within the full field of view, the overall correction model of the external orientation elements is established by connecting the points in the overlapping area, and the calibration parameters of multiple external orientation elements are multimodally fused to complete the consistency correction of the external orientation elements. Through the mutual iterative optimization of the internal and external orientation elements, high-precision geometric calibration with a large field of view is completed.

[0027] An on-orbit geometric calibration method for a large-field-of-view optical remote sensing satellite based on multimodal fusion describes the above four steps in detail respectively.

[0028] Step 1 is the geometric calibration model adopted by the present invention, which includes the following steps:

[0029] The geometric calibration of single satellite observation data uses high-precision geometric calibration field data as a reference to build a geometric calibration model for cloud-free areas. By solving the geometric calibration model, the internal orientation elements and camera external orientation elements corresponding to each pixel in the sensor slice of a single cloud-free area of ​​camera imaging are obtained. The specific implementation method includes the following sub-steps:

[0030] Step 1.1, extract image feature points on the cloud-free sensor chip of a single image to be calibrated, and densely match them with high-precision geometric calibration field data to generate control point data;

[0031] Step 1.2, using the single-shot imaging state data to be calibrated, including imaging time information, ephemeris orbit data, attitude data, and the camera ground calibration installation matrix, etc., to construct a geometric calibration model based on the detector pointing angle;

[0032] The geometric calibration model based on the detector pointing angle is as follows:

[0033]

[0034] in:

[0035] [X S ,Y S ,Z S ] is the position of the satellite in the agreed geocentric coordinate system, that is, the orbital position at the control point (PX, PY, PZ);

[0036] [X G ,Y G ,Z G ] is the coordinate of the ground point corresponding to the pixel in the agreed geocentric coordinate system;

[0037] psiX and psiY are the angles between the unit vector of the principal optical axis of the camera pixel corresponding to the image and the X-axis and Y-axis of the satellite body coordinate system, respectively;

[0038] u—proportional factor;

[0039] M0 is the installation matrix of the satellite body coordinate system relative to the camera, which is obtained by ground measurement before the satellite is launched;

[0040] M1 is the satellite-to-orbit coordinate system rotation matrix, which is composed of the attitude angles measured on the satellite;

[0041] M2 is the rotation matrix from the lower orbit to the J2000.0 coordinate system, which is composed of the right ascension of the ascending node, the orbit inclination, and the argument.

[0042] M3 is the rotation matrix from J2000.0 to WGS84 coordinate system, which requires corrections for precession, nutation, Greenwich sidereal time and polar motion.

[0043] Step 1.3, using the control point data obtained in step 1.1 and the geometric calibration model based on the detector pointing angle, the geometric internal orientation element is solved;

[0044] Step 1.4, using the internal orientation elements obtained in step 1.3, the control point data and the geometric calibration model based on the detector pointing angle, iteratively solve the external orientation elements, so as to obtain the camera installation matrix and the pointing angle of each detector in the camera coordinate system, and realize the calibration of the internal and external orientation elements of the camera;

[0045] Step 1.5, the geometric calibration model based on the detector pointing angle adopted in the present invention includes three detailed models: detailed model 1 is that the external orientation elements are known, and only the internal orientation elements are calibrated; detailed model 2 is that the internal orientation elements are known, and only the external orientation elements are calibrated; detailed model 3 is that both the internal and external orientation elements are unknown, and the internal and external orientation elements are solved iteratively.

[0046] Step 2 is a key step of the present invention, and its purpose is to solve the problem of how to achieve the continuity and consistency of the internal orientation elements obtained by the two solutions when geometric calibration is carried out on the two imaging data, so as to achieve seamless splicing of the whole image and high-precision distortion correction. Figure 2 The process in the dotted box and Figure 3 Schematic diagram, the specific steps are as follows:

[0047] Geometric calibration of cloud-free satellite observation data containing two overlapping areas. Based on the internal orientation elements of the first calibration of the overlapping area, the geometric calibration model of single satellite observation data is used to calculate the camera external orientation elements corresponding to the second imaging. The specific implementation method includes the following sub-steps:

[0048] Step 2.1, if the first camera sensor imaging obtains cloud-free images including slices 7-10, the second camera sensor imaging obtains cloud-free images from slices 10-14, where slice 10 is the overlapping sensor of the two imagings; when the first imaging has completed the geometric calibration processing of step 1, the camera exterior orientation element M has been obtained. exter[1] and the inner element P of the 7th to 10th slices inner1[7-10] .

[0049] Step 2.2, with P inner1

[10] To input the internal orientation elements, the detailed model 2 in step 1.5 is used to calibrate the external orientation elements of the second imaging data to obtain the external orientation elements M corresponding to the second imaging.exter[2] .

[0050] Step 3: Use the exterior orientation element M of the second imaging obtained by step 2.2 exter[2] , the internal orientation element calibration is performed on the sensor slices that do not overlap in the second imaging. The specific implementation method includes the following sub-steps:

[0051] Step 3.1, using the cloud-free sensor slices (10-14 slices) of the second imaging to densely match with the high-precision geometric calibration field data to obtain control point data;

[0052] Step 3.2, using the second imaging state data, including imaging time information, ephemeris orbit data, attitude data, and the external orientation element M obtained by calibration in step 2.2 exter[2] etc., use the detailed model in step 1.5 to calibrate the internal orientation elements of the second imaging cloud-free sensor slices (slices 11-14);

[0053] At this time, the second camera sensor imaging of the 10th piece and the 11th-14th pieces use the same external orientation elements, which can ensure the continuity and consistency of the internal orientation elements of the 10-14 sensors, and the internal orientation elements of the 10th piece are also continuous and consistent with the first imaging cloud-free sensor pieces 7-9, thereby realizing the internal orientation element transmission based on two different imaging state data and achieving seamless stitching between image slices.

[0054] The purpose of step 4 is to use multiple imaging data to perform geometric calibration and obtain internal orientation elements that are continuous and consistent across the entire field of view. Figure 2 The specific steps of the process are as follows:

[0055] Step 4: Using the cloud-free overlap area between other secondary imaging data and the first or second imaging data, the method of step 2 is repeated to obtain the exterior orientation elements M corresponding to other secondary imaging data. exter[n] After all calibrations are completed, the internal orientation elements P of all sensor pieces in the full field of view are obtained. inner[1-m] .

[0056] Step 4 is a key step of the present invention, and its purpose is to solve the problem of how to obtain uniform high-precision exterior orientation elements after obtaining the interior orientation elements of all sensor pieces within the full field of view using multiple imaging data. The specific steps are as follows:

[0057] The same-name points are selected in the overlapping area of ​​multiple imaging sensor slices, and the calibrated full-field internal orientation elements and multiple imaging state parameters including each imaging time information, ephemeris orbit data, and attitude data are used to obtain the adjusted external orientation element parameters. The specific implementation method includes the following sub-steps:

[0058] Step 4.1, select the same-name points in the overlapping area of ​​multiple imaging slices, and use the calibrated full-field orientation elements and multiple imaging state parameters to establish the error equation:

[0059] For a set of overlapping points with the same name, the actual geographic coordinates are (lat, lon, height), and the coordinates converted to the geocentric coordinate system of the protocol are [X G ,Y G ,Z G ]; the coordinates of this point on the first image image1 are (sample1, line1), and the transformation matrix is ​​M exter[1] , the coordinates on the second image image2 are (sample2, line2), and the transformation matrix is ​​M exter[2] , with P inner Substitute (psiX, psiY) as the true value into the geometric calibration model based on the detector pointing angle, and use M exter For the unknown parameters to be solved, establish the error equation.

[0060] make

[0061] Among them, a1, b1, c1, a2, b2, c2, a3, b3, c3 are the 9 parameters of the exterior orientation elements.

[0062] Then the following error equation can be established:

[0063]

[0064] Multiple groups of overlapping points P inner (psiX1, psiY1), P inner (psiX2, psiY2)..., establish a set of equations and construct an overall correction model for the external orientation elements.

[0065]

[0066] Step 4.2, based on the least squares adjustment solution, obtain the adjusted exterior orientation elements a1, b1, c1, a2, b2, c2, a3, b3, c3.

[0067] The internal orientation elements obtained in step 3.2 and the external orientation elements obtained in step 5.2 are the on-orbit geometric calibration parameters of the large-field-of-view optical remote sensing satellite based on multi-modal fusion.

[0068] The above contents are further detailed descriptions of the present invention in combination with specific implementation methods, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A geometric calibration method for large-field-of-view optical remote sensing satellite based on multi-state data fusion, characterized in that: The steps include: Step 1: geometric calibration of single satellite observation data. Using high-precision geometric calibration field data as a reference, a geometric calibration model for cloud-free areas is constructed. By solving the geometric calibration model, the internal orientation elements and camera external orientation elements corresponding to each pixel in the sensor of a single cloud-free area of ​​camera imaging are obtained. Step 2: Geometric calibration of cloud-free satellite observation data containing two overlapping areas. Based on the internal orientation elements of the first calibration of the overlapping area, the geometric calibration model of single satellite observation data is used to calculate the camera exterior orientation elements corresponding to the second imaging sensor; then the camera exterior orientation elements are used to calibrate the internal orientation elements of the sensor without overlap in the second imaging; Step 3, using the overlapping area between other secondary imaging data and the first or second imaging data, and using the method of step 2, obtaining the exterior orientation elements of other secondary imaging data and the interior orientation elements of all other sensors; Step 4: Based on the acquired internal orientation elements within the full field of view, the overall correction model of the external orientation elements is established by connecting the points in the overlapping areas, and the calibration parameters of multiple external orientation elements are multi-modally fused to complete the consistency correction of the external orientation elements; the high-precision geometric calibration with a large field of view is completed by iteratively optimizing the internal and external orientation elements. Step 4.1, select the same-name points in the overlapping area of ​​multiple imaging slices, and use the calibrated full-field orientation elements and multiple imaging state parameters to establish the error equation: For a set of overlapping points with the same name, the actual geographic coordinates are (lat, lon, height), and the coordinates converted to the geocentric coordinate system of the protocol are [X G ,Y G ,Z G ]; the coordinates of this point on the first image image1 are (sample1, line1), and the transformation matrix is ​​M exter[1] , the coordinates on the second image image2 are (sample2, line2), and the transformation matrix is ​​M exter[2] , with P inner Substitute (psiX, psiY) as the true value into the geometric calibration model based on the detector pointing angle, and use M exter Establish error equations for unknown parameters to be determined; make Among them, a1, b1, c1, a2, b2, c2, a3, b3, c3 are 9 parameters of the exterior orientation elements; Then the following error equation can be established: Multiple groups of overlapping points P inner (psiX1, psiY1), P inner (psiX2, psiY2)..., establish a set of equations and construct an overall correction model for the external orientation elements. Step 4.2, based on the least squares adjustment solution, obtain the adjusted exterior orientation elements a1, b1, c1, a2, b2, c2, a3, b3, c3; The internal orientation elements obtained in step 3.2 and the external orientation elements obtained in step 5.2 are the on-orbit geometric calibration parameters of the large-field-of-view optical remote sensing satellite based on multi-modal fusion.

2. The method for geometric calibration of optical remote sensing satellites with a large field of view based on multi-state data fusion according to claim 1, characterized in that: In step 1, the geometric calibration of single satellite observation data is carried out by building a geometric calibration model for cloud-free areas with reference to high-precision geometric calibration field data. The internal orientation elements and camera external orientation elements corresponding to each pixel in the sensor slice of a single cloud-free area imaged by the camera are obtained by solving the geometric calibration model. The specific implementation method includes the following sub-steps: Step 1.1, extract image feature points on the cloud-free sensor chip of a single image to be calibrated, and densely match them with high-precision geometric calibration field data to generate control point data; Step 1.2, using the single-shot imaging state data to be calibrated, including imaging time information, ephemeris orbit data, attitude data, and the camera ground calibration installation matrix, etc., to construct a geometric calibration model based on the detector pointing angle; Step 1.3, using the control point data obtained in step 1.1 and the geometric calibration model based on the detector pointing angle, the geometric internal orientation element is solved; Step 1.4, using the internal orientation elements obtained in step 1.3, the control point data and the geometric calibration model based on the detector pointing angle, iteratively solve the external orientation elements, so as to obtain the camera installation matrix and the pointing angle of each detector in the camera coordinate system, and realize the calibration of the internal and external orientation elements of the camera; In step 1.5, the geometric calibration model based on the detector pointing angle includes three detailed models: Detailed model 1 has known external orientation elements, and only the internal orientation elements are calibrated; Detailed model 2 has known internal orientation elements, and only the external orientation elements are calibrated; In the detailed model 3, both the internal and external orientation elements are unknown, and the internal and external orientation elements are solved iteratively.

3. The method for geometric calibration of optical remote sensing satellite with large field of view based on multi-state data fusion according to claim 1, characterized in that: The step 2 specifically includes the following steps: Step 2.1, if the first camera sensor imaging obtains cloud-free images including slices 7-10, the second camera sensor imaging obtains cloud-free images from slices 10-14, where slice 10 is the overlapping sensor of the two imagings; when the first imaging has completed the geometric calibration processing of step 1, the camera exterior orientation element M has been obtained. exter[1] and the inner element P of the 7th to 10th slices inner1[7-10] ; Step 2.2, with P inner1[10] To input the internal orientation elements, the detailed model 2 in step 1.5 is used to calibrate the external orientation elements of the second imaging data to obtain the external orientation elements M corresponding to the second imaging. exter[2]。 4. The method for geometric calibration of optical remote sensing satellite with large field of view based on multi-state data fusion according to claim 2, characterized in that: In step 3, the exterior orientation element M of the second imaging obtained by solving step 2.2 is used. exter[2] , the internal orientation element calibration is performed on the sensor slices that are not overlapped in the second imaging, and the specific implementation method includes the following sub-steps: Step 3.1, using 10-14 cloud-free sensor slices from the second imaging, densely matching with the high-precision geometric calibration field data to obtain control point data; Step 3.2, using the second imaging state data, including imaging time information, ephemeris orbit data, attitude data, and the external orientation element M obtained by calibration in step 2.2 exter[2] etc., using the detailed model in step 1.5 to calibrate the internal orientation elements of the second imaging cloud-free sensor slices 11-14; At this time, the second camera sensor imaging of the 10th piece and the 11th-14th pieces use the same external orientation elements, which can ensure the continuity and consistency of the internal orientation elements of the 10-14 sensors, and the internal orientation elements of the 10th piece are also continuous and consistent with the first imaging cloud-free sensor pieces 7-9, thereby realizing the internal orientation element transmission based on two different imaging state data and achieving seamless stitching between image slices.

Images